Synergistic Effects of Acyclic Retinoid and OSI-461 on Growth Inhibition and Gene Expression in Human Hepatoma Cells

Hepatoma (hepatocellular carcinoma) is the fifth most frequent malignancy worldwide with an annual incidence of 560,000 cases (1). The development of hepatoma is frequently associated with chronic inflammation of the liver induced by persistent infection with hepatitis B virus or hepatitis C virus. The prognosis for patients with hepatoma is poor because there is no effective treatment of metastatic disease. Furthermore, even in early stage cases where surgical treatment might be expected to be curative, the incidence of recurrence in patients with underlying cirrhosis is ∼20 to 25% a year (2, 3). In addition, at present there are no effective chemopreventive agents for hepatoma. Therefore, there is a critical need to develop more effective strategies for the chemoprevention and chemotherapy of hepatoma.

Retinoids, a group of structural and functional analogs of vitamin A, exert fundamental effects on the regulation of epithelial cell growth, differentiation, and development (4). Some retinoids have been shown to have chemopreventive and chemotherapeutic activity for various human cancers (5). Acyclic retinoid, a novel synthetic retinoid, inhibits experimental liver carcinogenesis and induces apoptosis in human hepatoma-derived cells (6, 7, 8). A clinical trial showed that the administration of acyclic retinoid reduced the incidence of post-therapeutic recurrence of hepatoma and improved the survival rate of patients (9, 10). These favorable effects were associated with minimal toxicity. We reported recently that in a human hepatoma cell line acyclic retinoid causes an arrest of the cell cycle in G₀-G₁, increased cellular levels of the p21^CIP1 protein, and decreased levels of the cyclin D1 protein (11). It has been reported that acyclic retinoid acts synergistically with IFN in suppressing the growth and inducing apoptosis in human hepatoma-derived cells (12). Therefore, acyclic retinoid may be a valuable agent in the chemoprevention and chemotherapy of hepatoma, and its efficacy may be enhanced by combination with agents that target other signaling pathway in hepatoma cells.

Retinoids are thought to exert their biological function by regulating gene expression primary through two distinct nuclear receptors, retinoic acid receptors (RARs) and retinoid X receptors (RXRs), both composed of three subtypes (α, β, and γ; ref. 13). RARs modulate the expression of target genes by interacting with the retinoic acid response element (RARE) located in the promoter regions of target genes (13). Both RARs and RXRs can play important roles in influencing the growth of various cancers including hepatoma. Thus, RXRα binds to the enhancer element of hepatitis B virus (14). In addition, phosphorylation of RXRα occurs in hepatoma, and this impairs the function of this receptor (15, 16). The RARα gene is located near one of the integration sites of hepatitis B virus, and its expression is induced in hepatoma (17). The RARβ gene can also be an integration site of hepatitis B virus (18). The expression of RARβ is markedly decreased in human hepatoma and in human and rat hepatoma-derived cell lines (19, 20). Among the retinoid receptors, RARβ is thought to be one of the most important receptors with respect to regulation of cell growth and apoptosis. Inactivation of the RARβ gene abrogates retinoic acid-induced growth arrest (21), and the induction of RARβ expression is associated with strong enhancement of retinoic acid-induced growth inhibition in breast carcinoma and squamous cell carcinoma cell lines (22, 23). Treatment of human hepatoma cells with all-trans-retinoic acid causes a marked increase in cellular levels RARβ mRNA but not RARα mRNA (24). Acyclic retinoid also causes an increase in cellular levels of RARβ mRNA, but not RXRα mRNA, and suppresses expression of the c-myc oncogene mRNA in PLC/PRF/5 and Huh7 human hepatoma cells (25). We found that acyclic retinoid specifically caused an increase in the levels of expression of the RARβ protein and mRNA in human esophageal squamous cell carcinoma cells (26). These observations suggest that RARβ is an important mediator of the action of acyclic retinoid and related retinoids, and that induction of the expression of RARβ by acyclic retinoid contributes to the inhibition of growth of tumor cells, including hepatoma cells.

OSI-461 is potent derivative of Apotosyn (sulindac sulfone), which inhibits the cyclic GMP-specific phosphodiesterases 2 and 5 (27). This results in persistent increases in cellular levels of cyclic GMP, which leads to activation of protein kinase G and thereby the induction of apoptosis (27). In colon cancer cells, this process is associated with activation of c-Jun NH₂-terminal kinase 1 and reduction of cellular levels of β-catenin and cyclin D1 (27, 28, 29). In chronic lymphocytic leukemia cells, OSI-461 causes an arrest of the cell cycle in G₂-M, induces phosphorylation of Bcl-2 protein, and induces apoptosis (30). OSI-461 can also induce G₂-M arrest in colon cancer cells,⁴ but the mechanism responsible for this effect is not known. Thus, OSI-461 can exert growth inhibitory effects in a variety of human cancer cells, apparently via diverse mechanisms. However, there are no previous detailed studies on the anticancer effects of OSI-461 or related compounds in hepatoma cells.

The combined use of two or more chemotherapy agents is often advantageous as it may permit lowering of drug dosages and consequently decrease toxicity, reduce the opportunity for the development of drug resistance by cancer cells, and provide the potential for synergistic effects between drugs that act with different mechanisms. There is growing interest in increasing the antitumor activity of retinoids by combination with other agents, including a histone deacetylase inhibitor (31), IFN-β (12), the tumor necrosis factor-related apoptosis-inducing ligand (32), and conventional chemotherapy agents (33). The synergistic effect of the combination of a retinoid with a histone deacetylase inhibitor appears to be because of enhancement of the retinoid receptor signaling pathway (31). However, the molecular mechanisms responsible for the synergistic effects observed with combinations of retinoids with other anticancer agents are not known. The purpose of this study was to examine whether the combination of acyclic retinoid plus OSI-461 exerts synergistic growth inhibitory effects on human hepatoma cells and to examine possible mechanisms for such synergy, focusing on retinoid receptors and cell cycle control molecules.

Acyclic retinoid (NIK333, Nikken Chemicals Co., Ltd., Tokyo, Japan) was provided by Dr. Hisataka Moriwaki (Gifu University School of Medicine, Gifu, Japan; ref. 6). OSI-461 was supplied by OSI Pharmaceuticals, Inc. (Farmingdale, NY).

The HepG2 human hepatoma cell line was obtained from the American Type Culture Collection (Manassas, VA) and was maintained in DMEM supplemented with 10% fetal bovine serum (DF10; Invitrogen, Carlsbad, CA). Cells were cultured in an incubator with humidified air with 5% CO₂ at 37°C.

Cell proliferation was measured by a colony formation assay as described previously (11, 34). Three thousands cells of HepG2 cells were plated into 6 cm-diameter culture dishes. The following day, acyclic retinoid and OSI-461 were added alone or in combination at the indicated concentrations to each plate, and the cells were incubated for an additional 5 days. The colonies were then fixed with 70% EtOH, stained with Giemsa solution, and counted. Results were expressed as % colony formation with 100% representing control cells treated with 0.1% DMSO alone. All assays were done in triplicate, and mean values were plotted. To determine whether the combined effects of acyclic retinoid plus OSI-461 were synergistic, the combination index-isobologram was used in the drug combination assays (35).

We used DNA fragmentation and caspase activity assays to detect Apoptosis. HepG2 cells were treated with DMSO (Group 1), 5 μmol/L acyclic retinoid alone (Group 2), 0.1 μmol/L OSI-461 alone (Group 3), or the combination of 5 μmol/L acyclic retinoid plus 0.1 μmol/L OSI-461 (Group 4), respectively, for the indicated time (0, 24, 48, and 72 hours), and the cell lysates were used for apoptosis assays. Briefly, the DNA fragmentation assay was done with the Cell Death Detection ELISA^PLUS kit (Roche Diagnostics Co., Indianapolis, IN), which detects the presence of histone-associated DNA fragmentation in the cell cytosol, according to the manufacturer’s instructions. The levels of DNA fragmentation found in Group 1 at 0 hours was set to 1, and changes were shown as fold activation. Caspase activity assays were done with a fluorometric system as described previously (36). The cell extracts (10 μg of total protein) were incubated with 2 μg of the fluorogenic peptide substrates Ac-DEVD-AFC, Ac-IETD-AFC (PharMingen, San Diego, CA), or Ac-LEHD-AFC (EMD Biosciences Inc., San Diego, CA), the fluorogenic substrates for caspase-3, -8, and -9, respectively, and caspase activities were calculated by release of AFC, which was measured with a Gemini Fluoro/Luminometer (Molecular Devices Co., Sunnyvale, CA). The caspase activities found in Group 1 at each time point were set to 1, and changes were shown as fold activation.

HepG2 cells were treated with test drugs for the indicated time (0, 24, 48, and 72 hours). The harvested cells were stained with propidium iodide (Sigma, St. Louis, MO), and the samples were then analyzed for DNA histograms and cell cycle phase distribution by flow cytometer with a FACS Calibur instrument (Becton Dickinson, Franklin Lakes, NJ), and the data were analyzed with the CELL Quest computer program (Becton Dickinson), as described previously (11, 26, 34).

Intracellular levels of cyclic GMP assays were done as described previously (37). HepG2 cells were plated at 1 × 10⁶ cells/10-cm-diameter plates, and on the following day the test drugs were added. After 1 hour of incubation, the cells were lysed, and the cell extracts were assayed for intracellular levels of cyclic GMP with an enzyme-immunoassay system (cGMP EIA system; Amersham Biosciences, Piscataway, NJ) according to the manufacturer’s instructions.

HepG2 cells were treated with the four conditions described above for the indicated time (0, 3, 6, 12, 24, and 48 hours). Protein extracts were then prepared as described previously (11, 26). The cell lysates (20 to 80 μg/lane) were separated by SDS-PAGE with 7.5 to 15% polyacrylamide gels. The primary antibodies for cyclin D1, p16^INK4A, p21^CIP1, p27^KIP1, cdk4, cdk6, ppRb, Bax, Bcl-2, Bcl-x_L, β-catenin, RARα, RARβ, RARγ, RXRα, RXRβ, RXRγ, vasodilator-stimulated phosphoprotein (VASP), and actin were obtained as described previously (11, 26, 34, 37). The p53 (C-19) and Apaf-1 (H324) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). Antimouse IgG, antirabbit IgG (Amersham Biosciences), or antigoat IgG (Santa Cruz Biotechnology) antibodies were used as the secondary antibodies. Each membrane was developed with an ECL-enhanced chemiluminescence system (Amersham Biosciences). An antibody to actin served as a loading control. The intensities of blots were quantified with NIH Image software version 1.62. The intensities of blots found at the 0-hour time point in each Group was set to 1, and changes in expression were shown as the fold difference.

RNA extraction and semiquantitative reverse transcription-PCR analysis were done as described previously (11, 26). Total RNA was isolated from frozen HepG2 cells with TRIzol reagent as recommended by the manufacturer (Invitrogen). The cDNA was amplified from 1 μg of total RNA with SuperScript one-step reverse transcription-PCR with the platinum Taq system (Invitrogen). PCR was conducted for 25 to 32 cycles in a thermal controller (Programmable Thermal Controller; MJ Research Inc., Watertown, MA). Cyclin D1, p21^CIP1, RARβ, and actin-specific primer sets and the amplification cycle condition were as described previously (11, 26). Amplified PCR products obtained with actin-specific primers served as internal controls.

Reporter assays were done as described previously (38). The cyclin D1 promoter luciferase reporter plasmid-1745CD1LUC was kindly provided by Dr. Richard G. Pestell (Georgetown University, Washington, DC; ref. 39). The p21^CIP1 promoter luciferase reporter plasmid WWP-Luc was kindly provided by Dr. Kenneth W. Kinzler and Dr. Bert Vogelstein (The Johns Hopkins University Medical Institutions, Baltimore, MD; ref. 40). One microgram of DNA of the cyclin D1 or p21^CIP1 luciferase reporter plasmids were transfected into HepG2 cells (1 × 10⁵) with Lipofectin reagent (Invitrogen) in opti-MEM I medium (Invitrogen). After 24 hours the medium was changed to DF10 medium containing the specific drugs indicated in the figure legends. The cells were then incubated for indicated times, and luciferase activities were determined with a luciferase assay system (Program Co., Madison, WI). The cells were also cotransfected with a cytomegalovirus-β-galactosidase reporter. Luciferase activities were normalized with respect to the β-galactosidase activities, to correct for differences in transfection efficiency.

The RARE-CAT reporter plasmid was kindly provided by Dr. David A. Talmage (Columbia University, New York, NY; ref. 41). Transient transfection assays with a RARE promoter reporter were done as described above. Two micrograms of a RARE-CAT plasmid was transfected into HepG2. The cells were treated with drugs for 24 hours, and CAT activity was measured with an ELISA kit (CAT ELISA, Roche Diagnostics Co.) according to the manufacturer’s instructions, as described previously (26). RARE-CAT activities were normalized with respect to the β-galactosidase activities, as described above.

Statistical analyses of DNA fragmentation assays, caspase activity assays, cyclic GMP assays, cyclin D1 reporter assays, p21^CIP1 reporter assays, and RARE-CAT assays were analyzed by either Student’s or Welch’s t test. The results were considered statistically significant if P was less than 0.05.

We found that acyclic retinoid and OSI-461 inhibited growth of the HepG2 cells with IC₅₀ values of about 30.5 and 0.65 μmol/L, respectively, when the cells were grown in DF10 medium (Fig. 1, A and B). We then examined the effects of combined treatment with a range of concentrations of both agents on the growth of HepG2 cells. We found that the combination of as little as 1.0 μmol/L acyclic retinoid and 0.01 μmol/L OSI-461 (about the IC₁₀ value for both compounds) exerted synergistic growth inhibition. Thus, when analyzed by the isobologram method (35), the combination index for 1.0 μmol/L acyclic retinoid plus 0.01 μmol/L OSI-461 was 2+ and for 1.0 μmol/L acyclic retinoid plus 1.0 μmol/L OSI-461 was 4+ (Fig. 1,C; Table 1). For subsequent mechanistic studies on the combined effects of acyclic retinoid plus OSI-461, we defined the following 4 conditions of treatment: (1) Group 1, control (0.1% DMSO-solvent); (2) Group 2, 5 μmol/L (about IC₁₅ value) acyclic retinoid alone; (3) Group 3, 0.1 μmol/L (about IC₃₂ value) OSI-461 alone; and (4) Group 4, the combination of 5 μmol/L acyclic retinoid plus 0.1 μmol/L OSI-461.

Because both acyclic retinoid (7, 8, 26) and OSI-461 (27, 28) induce apoptosis in cancer cells, we then examined whether the synergistic effect on growth inhibition we observed with combined treatment with these agents (Fig. 1; Table 1) also applied to the induction of apoptosis. To quantify the induction of apoptosis, we measured histone-associated DNA fragmentation with an ELISA system (Fig. 2,A). We found that the combination of 5 μmol/L acyclic retinoid plus 0.1 μmol/L OSI-461 (Group 4) caused a strong time-dependent induction of apoptosis; the DNA fragmentation ratio increased by 1.6-fold after 24 hours and peaked at 3.8-fold after 48 hours. In contrast, no substantial changes were observed in Groups 2 (5 μmol/L acyclic retinoid alone) or 3 (0.1 μmol/L OSI-461 alone) treated cells during this time course (Fig. 2,A). We then examined the cellular levels of four apoptosis-related proteins by Western blot analysis (Fig. 2,B). We found that in Group 4 there was an increase in the levels of expression of the pro-apoptotic proteins Bax and Apaf-1 and a decrease in the levels of the anti-apoptotic proteins Bcl-2 and Bcl-x_L (Fig. 2,B). In caspase activity assays (Fig. 2,C), the Group 4 extracts revealed a substantial and time-dependent increase in caspase-9 and -3 activities, which occurred within 24 hours of addition of the two drugs. Caspase-8 activity was also substantially increased after 48 hours (Fig. 2,C). In contrast, no substantial changes were observed in the levels of apoptosis-related proteins (Fig. 2,B) or caspase activities (Fig. 2 C) in the Groups 2 or 3 extracts.

Previous studies showed that acyclic retinoid can induce G₀-G₁ cell cycle arrest in HepG2 cells (11) and that OSI-461 can induce G₂-M cell cycle arrest in chronic lymphocytic leukemia cells (30). To determine whether the synergistic effects on growth inhibition (Fig. 1; Table 1) and induction of apoptosis (Fig. 2,A) caused by combined treatment with these agents were associated with specific changes in cell cycle distribution, cell cycle analysis was done with DNA flow cytometry. The data are summarized in Fig. 2,D. With the combined treatment (Group 4), the percentage of cells in G₀-G₁ increased by 8.5% after 24 hours and by 17.7% after 48 hours when compared with control untreated cells, and this was associated with a concomitant decrease of cells in the S phase of the cell cycle. Moreover, at 72 hours, there was a marked increase in the sub-G₁ fraction, which is a marker of apoptosis, thus confirming the results on apoptosis obtained in Fig. 2,A. When compared with the control (Group 1), the Group 2 cells but not the Group 3 cells also displayed an increase in G₀-G₁ cells at 48 hours, but neither group displayed an increase in sub-G₁ cells (Fig. 2 D).

As discussed in the Introduction section, OSI-461 can induce apoptosis in colon cancer cells through a cyclic GMP/protein kinase G pathway (27, 28). To determine whether the induction of apoptosis in HepG2 cells caused by the synergistic combination of acyclic retinoid plus OSI-461 (Fig. 2,A) was associated with activation of this pathway, we first measured intracellular levels of cyclic GMP (Fig. 3,A). As a positive control we treated the cells with a relatively high concentration of OSI-461, i.e., 10 μmol/L. As expected from previous studies (37), at 1 hour this substantially increased the intracellular level of cyclic GMP. However, there were no substantial increases in cyclic GMP in Groups 1, 2, 3, or 4 (Fig. 3,A). We then examined the expression level of the hyperphosphorylated form of the vasodilator-stimulated phosphoprotein protein (VASP-P) because VASP phosphorylation provides a convenient biomarker for the activation of protein kinase G in tumor cells (37). Treatment of HepG2 cells with 10 μmol/L OSI-461 for 1 hour caused a marked increase in VASP-P, but no substantial increase was seen in the extracts from the Groups 1, 2, 3, or 4 cells (Fig. 3,B). When we examined the time-dependent effect of acyclic retinoid and OSI-461 on cellular levels of VASP-P, across a time period of 0 to 48 hours, there was also no substantial change in Groups 1, 2, 3, or 4 (Fig. 3,C). Thus, it appears that the growth inhibition (Fig. 1) and apoptosis (Fig. 2 A) induced by the combination of only 0.1 μmol/L OSI-461 and 5 μmol/L acyclic retinoid (Group 4) is not associated with detectable activation of protein kinase G (see Discussion section).

Because treatment of HepG2 cells with acyclic retinoid plus OSI-461 caused an increase of cells in the G₀-G₁ phase of the cell cycle (Fig. 2,D), we examined by Western blot analysis the effects of this combination on cellular levels of the G₁ cell cycle control proteins cyclin D1, cdk4, cdk6, and the cyclin-dependent kinase inhibitors p16^INK4a and p27^KIP1 (Fig. 4, A and D). We also examined cellular levels of the β-catenin protein because its accumulation transactivates Lef/Tcf transcription, thus enhancing the expression of cyclin D1 (42), and OSI-461 can cause a decrease in cellular levels of both the cyclin D1 and β-catenin proteins in SW480 human colon cancer cells (29). We found that the level of cyclin D1 protein displayed a moderate decrease at 24 hours, and a marked decrease at 48 hours in the combined treatment Group 4 cells. There was also a moderate decrease in level of the cyclin D1 protein at 48 hours in Group 2, but no substantial change in cells treated with only 0.1 μmol/L OSI-461 (Group 3). No appreciable changes in the cellular levels of β-catenin protein were found in Groups 1, 2, 3, or 4 at these time points (Fig. 4,A). The decreases in cyclin D1 protein were paralleled by similar decreases in its mRNA (Fig. 4,B). In addition, in transient transfection reporter assays (Fig. 4,C) there were substantial decreases in cyclin D1 promoter activity in Group 4, especially, in assays done for 48 hours. Cells treated with 5 μmol/L acyclic retinoid alone (Group 2) also displayed a substantial decrease in cyclin D1 promoter at 48 hours, but this effect was less than that seen with the combination of drugs. No substantial inhibition was seen in cells treated with only 0.1 μmol/L OSI-461 (Fig. 4,C). Taken together, with the above studies on mRNA levels (Fig. 4 B) these results provide evidence that the combined inhibitory effects of acyclic retinoid plus OSI-461 on expression of cyclin D1 are exerted at the level of transcription of the cyclin D1 gene.

The tumor suppressor protein retinoblastoma protein (pRb) inhibits the G₁ to S transition, but when it becomes hyperphosphorylated (designated ppRb) it no longer exerts this inhibitory effect (43, 44). Therefore, we also examined whether treatment with these agents alters the cellular level of ppRb (Fig. 4,D). We found that the level of ppRb displayed a marked decrease in Group 2 cells and was virtually undetectable in Group 4 cells at 48 hours after addition of the drugs. However, no change was observed in the Group 3 cells. No detectable change was seen in cellular levels of the cdk4, cdk6, p16^INK4a, or p27^KIP1 proteins in Groups 1, 2, 3, or 4 (Fig. 4 D).

In view of our finding that the combination of acyclic retinoid plus OSI-461 caused a decrease in the expression of cyclin D1 and ppRb (Fig. 4) and an increase of cells in the G₀-G₁ phase of the cell cycle (Fig. 2,D), we then examined cellular levels of p21^CIP1 because it binds to cyclin D1-cdk4/6 complexes, thus inhibiting phosphorylation of pRb and delaying transition from the G₁ phase to the S phase of the cell cycle (Fig. 5, A–C; refs. 45, 46). We also examined cellular levels of p53 because activation of p53 induces the expression of p21^CIP1 (Fig. 5,D; ref. 40). When we examined the effects of acyclic retinoid and OSI-461 on cellular levels of p21^CIP1 during a 48 hours treatment period, there was a marked increase in the p21^CIP1 protein at 6 to 12 hours in both Groups 2 and 3 cells. This effect was even stronger in the Group 4 cells, and the increase appeared within 3 hours and persisted for at least 48 hours (Fig. 5,A, top and bottom panels). Somewhat similar effects were seen with respect to levels of p21^CIP1 mRNA, when determined by reverse transcription-PCR analysis (Fig. 5,B). In transient transfection reporter assays (Fig. 5,C), we also found that 5 μmol/L acyclic retinoid alone (Group 2) and 0.1 μmol/L OSI-461 alone (Group 3) caused an increase in transcriptional activity of p21^CIP1 promoter when assayed at 6 hours. Moreover, the combination of these agents (Group 4) caused a synergistic increase; the relative luciferase activity increased by 3.9-fold compared with the Group 1 control cells (Fig. 5,C). Taken together, with the above studies on mRNA levels (Fig. 5,B) these results provide direct evidence that the p21^CIP1 gene was regulated transcriptionally by the combined treatment of acyclic retinoid plus OSI-461. When we examined time-dependent effects of these agents on cellular levels of the p53 protein, we found that there was an increase of this protein at 6 to 24 hours in Groups 3 and 4 cells. However, no substantial changes were observed in cells treated with 5 μmol/L acyclic retinoid alone (Group 2; Fig. 5 D, left and right panels).

Because retinoids are thought to exert most of their effects on regulating gene expression by binding to specific retinoid nuclear receptors (13), we then examined effects of the combination of acyclic retinoid plus OSI-461 on cellular levels of specific retinoid receptors (Fig. 6, A–C). We focused on RARβ because of our previous evidence, which suggests that acyclic retinoid acts mainly through this receptor (26). Fig. 6,A indicates that when we examined the time-dependent effect of 5 μmol/L acyclic retinoid (Group 2), we found a small increase in the levels of the RARβ protein after 6 hours, but no substantial effect of 0.1 μmol/L OSI-461 (Group 3). The combination of these two drugs (Group 4) produced a stronger and more persistent induction of RARβ than that seen with acyclic retinoid alone (Fig. 6,A, top and bottom panels). Qualitatively similar results were seen with respect to levels of RARβ mRNA in Groups 2 and 4. For reasons that are not apparent, the levels of RARβ appeared to decline in the Group 1 samples at 24 and 48 hours (Fig. 6,B). No appreciable changes were found in the levels of the other retinoid receptor proteins RARα, RARγ, RXRα, RXRβ, and RXRγ when assayed at 12 hours in the Groups 1 to 4 cells (Fig. 6 C).

RARs modulate the expression of target genes by interacting with RARE elements located in the promoter regions of target genes (13). In view of our finding that the combination of acyclic retinoid plus OSI-461 caused a rapid and marked increase in the levels of expression of both RARβ protein and mRNA (Fig. 6, A and B), we then examined whether the combination of these agents stimulates the transcriptional activity of a RARE-CAT reporter in transient transfection reporter assays (Fig. 6,D). We found that 30 μmol/L acyclic retinoid caused a marked (2.4-fold) increase of RARE reporter activity, whereas a high concentration of OSI-461 (10 μmol/L) had no effect. When HepG2 cells were treated with 5 μmol/L acyclic retinoid (Group 2) or 0.1 μmol/L OSI-461 (Group 3), there was no substantial stimulation. However, when these two treatments were combined (Group 4), there was a substantial 1.7-fold stimulation of receptor activity (Fig. 6 D).

In the present study, we found that the combination of low concentrations of acyclic retinoid and OSI-461 caused synergistic inhibition of growth of human hepatoma HepG2 cells and that this was associated with arrest of the cell cycle in G₀-G₁ and induction of apoptosis (Figs. 1 and 2; Group 4). In previous studies, we found that when acyclic retinoid was tested alone at its IC₄₀ concentration (20 μmol/L) it markedly inhibited growth of HepG2 cells and that this inhibition was associated with arrest of the cell cycle in G₀-G₁, increased cellular levels of the p21^CIP1 protein, decreased levels of cyclin D1, and inhibition of the transcriptional activity of the cyclin D1 promoter (11). We also found that acyclic retinoid inhibits the growth of esophageal squamous cell carcinoma cells and that this inhibition is associated with an increase in cellular levels of RARβ (26). Presumably, the latter effect enhances the expression of retinoid responsive genes and inhibition of AP-1 activity, thus contributing to the growth inhibitory effects of acyclic retinoid (26). In the present study, the combination of low concentrations of acyclic retinoid and OSI-461 had a synergistic effect in increasing cellular levels of RARβ and p21^CIP1, decreasing the expression of cyclin D1, and decreasing the level of ppRb (Figs. 4,5,6). This drug combination also enhanced induction of the pro-apoptotic proteins Bax and Apaf-1, reduced the levels of the antiapoptotic proteins Bcl-2 and Bcl-x_L, and caused activation of caspases-3, -8, and -9 (Fig. 2, B and C). Presumably, the latter effects play a role in the induction of apoptosis (Fig. 2 A).

A major question posed by the above findings is the molecular mechanisms by which the combination of low concentrations of acyclic retinoid and OSI-461 induce these diverse cellular and biochemical effects. A hypothetical scheme that addresses this question is shown in Fig. 7. This scheme emphasizes positive feedback interactions between the expression of p21^CIP1 and RARβ, but it is likely that other mechanisms also play a role. We will first focus on the promoter region of the p21^CIP1 gene. Nuclear retinoid receptors are ligand-dependent transcription factors that bind to RARE elements in the promoter regions of retinoid responsive target genes (13). The promoter region of the p21^CIP1 gene contains a variety of positive cis-acting elements, including both a RARE and a p53-responsive element (47). Indeed, previous studies indicate that all-trans-retinoic acid induces transcription of the p21^CIP1 gene by enhancing binding of the retinoid receptor complex to this RARE in the p21^CIP1 gene (48). Presumably, our finding that 5 μmol/L acyclic retinoid alone induces expression of p21^CIP1 at the level of transcription in HepG2 cells is explained by this mechanism (Fig. 5, A–C). We found that 0.1 μmol/L OSI-461 alone caused an increase in cellular levels of both p21^CIP1 and p53 (Fig. 5). It is known that p53 itself is a potent transcriptional activator of p21^CIP1 expression (40). This combined effect may explain the rapid, marked, and sustained increase in the levels of expression of both p21^CIP1 protein and mRNA at the level of transcription that we observed after treating HepG2 cells with acyclic retinoid plus OSI-461 (Fig. 5, A–C). OSI-461 can also induce p21^CIP1 expression in other cell types (30), but the mechanism is not known. This aspect is additionally discussed below. We found a marked decrease in cellular levels of the hyperphosphorylated form of the pRb protein in the combined treatment Group 4 cells (Fig. 4,D). The pRb protein plays a key role as a negative regulator of the G₁-S transition of the cell cycle by binding to and inhibiting the transcription factor E2F (43, 44). Activation of cyclin D-cdk4 or -cdk6 complexes results in hyperphosphorylation of pRb, thus relieving this inhibitory activity (43, 44). The p21^CIP1 protein can bind to these cyclin-cdk complexes and inhibit their kinase activity (43, 44). Thus, the marked and sustained induction of p21^CIP1 by the combination of acyclic retinoid plus OSI-461 may contribute to the arrest in G₀-G₁ that we observed in the Group 4 treated cells (Fig. 2,D). The other factor that could contribute to this effect is the marked decrease in cyclin D1 expression in the Group 4 treated cells (Fig. 4, A–C). Previous studies indicate that when tested alone at higher concentrations both acyclic retinoid and OSI-461 can inhibit the expression of cyclin D1 (11, 26, 29), and in the present study a low concentration of acyclic retinoid caused some inhibition of the expression of cyclin D1 (Fig. 4, A–C). The precise mechanism by which these two drugs inhibit cyclin D1 expression is not known.

We will next consider the RARβ gene. The promoter region of this gene contains a RARE (24, 49), which explains why 5 μmol/L acyclic retinoid alone can cause a transient increase in cellular levels of both the RARβ protein and mRNA (Fig. 6, A and B). The transcription of retinoid responsive target genes is also controlled by co-activators (5). When a ligand binds to RAR-RXR heterodimers, co-activators recruit the histone acetyltransferases, cAMP-responsive element-binding protein (CBP) and/or p300 (5). The acetylation of histone H4 by histone acetyltransferases results in nucleosomal repulsion, chromatin decondensation, and enhanced transcription (5). The transcriptional activation by both cAMP-responsive element binding protein (CBP) and p300 can be stimulated by co-expression of p21^CIP1 (50). Furthermore, introduction of the p21^CIP1 gene into cells transcriptionally activates the upstream promoter region of the RARβ gene (51). Therefore, our finding that the combination of acyclic retinoid plus OSI-461 increased cellular levels of RARβ and RARE-CAT activity (Fig. 6, A, B, and D) may be due, at least in part, to stimulation of the activities of CPB and/or p300 by the increased cellular levels of p21^CIP1. This sustained induction of RARβ and its activation by the ligand acyclic retinoid could produce a positive feedback effect on the promoter region of both the RARβ and p21^CIP1 genes, thus enhancing growth inhibition in HepG2 cells. Our findings that the p21^CIP1 promoter activity was synergistically stimulated by the combined treatment of acyclic retinoid plus OSI-461 is consistent with our hypothesis of a positive feedback effect between the RARβ and p21^CIP1 genes. Overexpression of RARβ in squamous cell carcinoma cells can inhibit cell proliferation in a ligand independent manner, and this inhibition is associated with induction of p300 and CBP (23). These effects as well as other putative growth inhibitory effects of RARβ (21, 22) could additionally contribute to the growth inhibition observed in the present study.

HepG2 cells contain a wild-type p53 gene (52). The p53 protein plays a critical role in induction of apoptosis in tumor cells by inducing the expression of proteins involved in the mitochondrial death receptor (DR) independent pathway and also in the DR-dependent pathway (53, 54). The induction of apoptosis by acyclic retinoid plus OSI-461 was associated with increased levels of the pro-apoptotic proteins Bax and Apaf-1 and decreased levels of the anti-apoptotic proteins Bcl-2 and Bcl-x_L (Fig. 2, A and B). RARβ appears to be the most important retinoid receptor with respect to induction of apoptosis (21, 22, 23). Transfection of the RARβ gene into RARβ-negative breast cancer cells in the presence of retinoic acid induces apoptosis (22), and transfection of RARβ into human squamous cell carcinoma cells in the presence of 9-cis-retinoic acid induces apoptosis, and this is associated with increased cellular levels of Bax and Apaf-1 (23). Therefore, it seems likely that the marked and sustained induction of RARβ in HepG2 cells treated with acyclic retinoid plus OSI-461 (Fig. 6, A and B) plays an important role in inducing apoptosis in these cells, and that this effect is mediated, at least in part, through the mitochondrial DR-independent pathway. Because we found an increase in caspase-8 activity in HepG2 cells treated with acyclic retinoid plus OSI-461 (Fig. 2,C), these agents may also activate the DR-dependent pathway of apoptosis (55). OSI-461 could exert this effect through the induction of p53 (Fig. 5 B). Acyclic retinoid also increases caspase-8 activity in human hepatoma JHH-7 cells (8). Furthermore, Sun et al. (32, 56) found that another synthetic retinoid 6-[3-(1-adamantyl)-4-hydroxyphenyl]-2-naphthalene carboxylic acid (CD437) can increase the expression of Fas, DR4, and DR5 in human lung cancer cells. Therefore, in future studies it will be of interest to additionally examine the effects of acyclic retinoid plus OSI-461 on activating the DR-dependent pathway.

As mentioned in the Introduction section, there is evidence that OSI-461-induced growth inhibition and apoptosis involve activation of the cyclic GMP-dependent enzyme protein kinase G (27, 28). Our laboratory found that in colon cancer cells phosphorylation of the cellular protein VASP by protein kinase G provides a convenient endogenous marker of the activation of this pathway (37). Treatment of chronic lymphocytic leukemia (30) and colon cancer cells⁵ with OSI-461 can cause G₂-M arrest of the cell cycle, but the mechanism responsible for this effect is not known. In the present study we found that treatment of HepG2 cells with 10 μmol/L OSI-461 (about the IC₉₀ concentration) induced an increase in the cellular level of cyclic GMP (Fig. 3,A), an increase in VASP-P (Fig. 3,B), and G₂-M arrest of the cell cycle (data not shown). However, the low dose of OSI-461 (0.1 μmol/L), which acted synergistically with acyclic retinoid to induce growth inhibition and apoptosis (Figs. 1 and 2,A), when used in combination with 5 μmol/L acyclic retinoid in HepG2 cells did not cause a detectable increase in the cellular level of cyclic GMP (Fig. 3,A), an increase in VASP phosphorylation (Fig. 3, B and C), or G₂-M arrest (Fig. 2,D). As mentioned above, this treatment did, however, cause an increase in cellular levels of both p21^CIP1 and p53 (Fig. 5). Because previous studies indicate that OSI-461 also induces p21^CIP1 expression in p53-mutated chronic lymphocytic leukemia (30) and SW480 colon cancer cells⁶, the induction of p21^CIP1 by OSI-461 might occur via the p53-independent pathway. However, the precise mechanism by which a low concentration of OSI-461 induces the expression of p21^CIP1 and p53 remains to be determined. Because it apparently occurs via a cyclic GMP/protein kinase G independent pathway, it may involve a yet to be discovered relatively high affinity target for this drug.

We should emphasize that the hypothetical scheme shown in Fig. 7 and the above discussion on possible molecular mechanisms do not exclude other possible explanations for our findings on the synergistic effects produced by acyclic retinoid plus OSI-461 on growth inhibition and apoptosis in HepG2 cells. Thus, RXRs may also play roles in regulating the promoter activity of the p21^CIP1 gene because transcriptional activation of this gene is preferentially stimulated by RAR-RXR heterodimers compared with RAR-RAR homodimers (48). Nevertheless, our results suggest that this combination of drugs, or combinations of similar drugs, may be useful in the chemoprevention and/or therapy of hepatoma and possibly other types of cancer. Indeed, in recent unpublished studies we found that the combination of all-trans-retinoic acid plus OSI-461 exerts synergistic inhibition of the growth of head and neck squamous cell carcinoma cells. These findings suggest the combination of specific retinoids with OSI-461 might be an effective regimen for the chemoprevention and/or chemotherapy of various types of human malignancies.

We thank Drs. Hisataka Moriwaki, Masataka Okuno, and Rie Matsushima-Nishiwaki (Gifu University School of Medicine, Gifu, Japan) for valuable discussions and comments.